How to Know God Exists is mostly about mutants; all it has to say about turtles is that there are no transitional fossils linking them to their presumed ancestors, the cotylosaurs. This is no longer quite true; there is Odontochelys semistestacea, a toothed (unlike toothless modern turtles) proto-turtle with a plastron (belly armor) but no shell on its back (to be sure, it is possible that this represents the secondary loss of an already-evolved upper shell), but transitional fossils leading up to true turtles are still rare. This does not, of course, make those australopiths and early hominines, whales with hind legs and small heads, feathered theropods, etc. go away. As noted in the review of the previous chapter, the fossil record is demonstrably very incomplete; gaps are to be expected.
Ray also briefly discusses vestigial organs, dismissing them as evidence for evolution on the grounds that they represent a loss rather than a gain of information, and because it can never be demonstrated conclusively that a vestige has no function. But total lack of function is not, of course, part of the definition of "vestigial," and has not been since Darwin noted that there were organs that had apparently lost their primary function even while remaining serviceable for other functions. One does not need to show, e.g. that an ostrich's wings have no function at all to notice that they don't serve the usual purpose of bird wings: ostriches cannot fly. One does not need to show that the human plantaris tendon does nothing at all to show that it doesn't enable humans to clench their feet into fists, as its homolog in nonhuman apes does. And one must wonder, if humans don't share ancestry with other primates, why we even have a tendon homologous to one that enables them to clench their feet (or why we have muscles to erect our no-longer-existent fur, or why we have pseudogenes by the dozen associated with smelling abilities we don't have. It's not mere paucity of function that makes vestiges evidence, but rather their homology -- their detailed similarity not required for similarity of function -- to structures that serve very different functions in obviously similar species.
Most of the chapter is about why mutations do not enable one "kind" to evolve into another. Ray spends a surprising number of lines and quote-mines to argue that mutations are random (this is in connection with a misunderstanding of punctuated equilibria, which Ray apparently thinks means that environmental pressures can cause a lizard to lay an egg that hatches out into a robin). That's not how he puts it; he argues that mutations exist before the environmental conditions that select for or against them, and do not arise in response to need, but then, that's more or less what "random mutations" means. He offers no dissent from the view that mutations occur, that they occur frequently, and that a few of them are beneficial, and that different mutations are beneficial in different environments. So we may take Ray as conceding that mutations occur and can be, on rare occasions, beneficial.
Ray argues that mutations do not "add information," and that even if they do, this added information cannot accumulate to produce novel organs, structures, and abilities. Ray is not nearly so clear on what "added information" actually would be, or how one would determine or compare the information content of a gene as he is that whatever information is, mutations cannot create it.
If "added information" is an attribute of the genome itself, one might suppose that, e.g. gene duplication would count, at least if one of the copies subsequently mutates (so that the descendant has all the genes of its ancestor plus a new, different gene). Known sorts of mutations include single-nucleotide substitutions, insertions or deletions of a single nucleotide, duplications, deletions, or transpositions of sections of the genome, and duplication of the entire genome (polyploidy). In principle, a succession of such mutations could change any genome to any other genome; if some genomes contain "more information" than others, then logically, known sorts of mutations ought to be able to add "information" to the genome.
If "added information" is an attribute of the phenotype, the organism's anatomy and behavior, then one would suppose that, e.g. the (micro)evolution of cecal valves in the wall lizards stranded on Pod Mrcaru would count, or the evolution of multicellularity in Chlorella vulgaris in the lab when the unicellular organism was exposed to a predator. The lizards have valves in their guts that their ancestors did not; the single-celled organisms were no longer isolated cells, but formed part of an symmetric ball of such cells (and required changes in their cell membranes to enable them to stick together in a regular fashion).
Ah, but Ray notes (with supporting quote mines, including one from Francisco Ayala, who, given that he is famous as a critic of creationism and the intelligent design movement, presumably does not actually think that mutations cannot accumulate into new structures, although he may well believe that accumulation of small adaptions does not cause speciation) that even if mutations add information, this added information can't keep accumulating into a new structure. What use, he asks, quoting Stephen Gould, would two percent of a wing be? Ray describes a 2% wing as a tiny, useless stub sticking out of the side of a bird; surely such a thing could not help the bird fly.
Now, what is 2% of a wing? One might be inclined to start with coelurosaurs, and see that 2% wing in the arm of something like the down-covered forelimb of Sinosauropteryx: in this case, that two percent would be an insulated, grasping or climbing arm. It would not work as a wing, but it would work as a forelimb suitable for a warm-blooded predator. A 20% arm would be something with more complete feathers, used for display, to look larger and more threatening or to impress potential mates, and a 50% wing (say, something like the fore- and hind limbs of Microraptor) would be useful for gliding (analogous to the "half-formed wings" of flying squirrels).
But perhaps that's too far along; perhaps a better example of a 2% wing would be the boneless pectoral fins of early Devonian jawless fish: they didn't enable their possessors to fly, grab things, or even crawl, but they provided weak aid to steering and propulsion. Perhaps a better example of a 20% wing would be the lobe fins of something like Tiktaalik, able to scull along the bottom of shallow bodies of water or even to push itself up. We might view Sinosauropteryx as at least halfway to a wing, and so forth.
The point is that we shouldn't think, in viewing the origin of organs and parts, of some beakless, wingless, legless, heartless and lungless and eyeless version of a crow struggling to survive and fly while natural selection worked frantically on random mutations to enable it to live as a crow. One percent of an eye functioned for some near-microscopic animal that benefited from the mere ability to detect changes in local light levels. One percent of a lung functioned in fish that got most of their oxygen through gills and benefited from (but didn't strictly need) a tiny boost from oxygen gulped from the air. One percent of a wing functioned as something other than a wing. Small benefits can accumulate because the function of structures evolves as the populations that bear those structures evolves.